This paper was motivated by a request to review relative operations performance for various fabrication facilities within a leading Taiwanese semiconductor manufacturer. Performance evaluation is important but often controversial. To dispel the controversy, we propose a two-stage fabrication process model to systematically analyze metrics currently adopted, and show that the commonly used wafer-based indices are biased for operations performance. Instead, they should be decomposed into productivity, representing true operations performance, and manufacturability. We suggest the use of data envelopment analysis because of its confirmed linkages to other widely used productivity measures and its overall performance via relative comparisons. The case study illustrates how the two-stage model evaluates and analyzes real-world operations, and the empirical results show the drawbacks of conventional methods.